The Business of Decision Making
The Economics and Effects of Where Businesses Make Decisions
In even the most hierarchical, command and control focused business, decisions are made throughout the organization. The lowly worker in the most systemized, by-the-book, fast food franchise is constantly deciding if they really want to do things by the book, or if they would prefer to go their own way and face the consequences. The issue is not if decisions are made by all employees at all levels, but which decisions — and if by making those decisions where they are currently made, are we making the decisions that are the best for the business.
To investigate this, we need to look at the economics of decision making in a company.
In 2011, a paper was published which examined the decisions of judges on Israeli parole boards. The judge's job was to examine the cases of individuals requesting parole and decide whether or not to grant it. One would hope such a decision was based on law, reason and the merit of each case. However, the results showed something different. In the early morning, judges were 65% percent likely to grant parole to a prisoner. But as lunchtime neared, the rate at which parole was granted dropped — until it was nearly 0% in the final cases of the morning. After lunch, the rate returned to 65% but then declined again until the next break.
As a prisoner, you would have to hope to be tried early in the morning than just before lunch!
The paper concluded that judges started off the day fresh, but as they had to make more and more decisions, they became tired and became more and more likely to make the safe — default — choice of not releasing the prisoner. A break refreshed the judges and enabled them to begin to make difficult decisions once again.
This effect is known as decision fatigue. Making difficult decisions taxes our willpower. When we’ve made too many difficult decisions, we revert to making the easiest decision, not necessarily the best.
In many businesses — indeed in the hierarchical structures not just of business, but also the military and government, difficult decisions tend to be pushed up the chain to command, towards the CEO or towards the Commander in Chief. George W Bush was mocked for saying it, but his job was indeed to be ‘The Decider’ — not so much a figurehead, but a point to which all decisions nobody else felt able to make were eventually passed. The study of judges highlights the problem with this approach to ‘pass the buck’ decision making. What if a decision to go to war or not had to be made just before lunch — after a grueling morning of less important, but equally challenging, decisions? A president might find themselves making the wrong choice and leading the world into a situation they would be better off out of.
We all know this intuitively. There is a reason why we sleep on difficult decisions: It gives us a chance to overcome our decision fatigue and return to being effective, and make better decisions.
Consider then the CEO who has innumerable decisions foisted upon them. Are they likely to be making good decisions? And if the CEO delegates most decisions, are those under him likely to be effective? The answer is a resounding no. We need to minimize the number of decisions any individual in a company is forced to make, in order to ensure they make the best decisions when called upon to do so. The more important the decisions an individual has to make, the fewer decisions they should make overall.
Now think back to the individual worker who is also constantly deciding whether or not to follow the rules of the book. Are they too likely to become fatigued? The answer is yes — at least at first. But also think back to the judges: the issue was not that they stopped making decisions, but that they defaulted to the easiest — the default — decision of “lock ’em up”. Outside of the workplace, we have a name for these easiest decisions: habits. An employee without the incentive and willpower to make a decision will continue to do what they habitually do. It is in a companies interest to ensure that employees are trained and reinforced around making the right set of default decisions. Such habits and reinforcement on mass become what we describe as a culture.
1 Minimise the number of decisions every employee has to make
1.1 Minimise further the number of important decisions any employee has to make
1.2 To avoid bad decisions being made, invest in building the correct culture
Imagine two companies. They are about to go head to head against one another to create the world's greatest widget. But each company has a different culture.
At the first company, they look at the problems that widgets can solve. They recruit the finest academics from the nearby university’s prestigious department of widgetology. They arrange a world-beating team of sales, marketing, and engineering leads. And they join one another in a conference room to determine their plan of attack. They plan carefully. They want no surprises. They spend weeks and months determining exactly what widgets the world needs, how to engineer the optimal widget with the least waste and use carefully crafter formulae from the foremost economic experts to price them. They go to the market with the perfect plan.
Meanwhile, the other company begins in a basement with far fewer employees. One is a dropout from a community college widgetology program, and another has a history of selling cheap Chinese widgets through the Amazon marketplace. They aim to get their custom-designed widget to market as quickly as possible. At first, the widget is expensive and their margins are low. They get a few buyers, but not the number they expected. So they iterate. They change the design of the widget to make it less good, but cheaper to manufacture. They experiment with a different manufacturing technique to see if they can cut costs. Each experiment is small — and most of the experiments fail — but a few successes but might add a couple of extra users, or get some extra pennies into the margin. And as more successes accumulate, each small percentage increase compounds. By the time the first company launches its world-beating product, the second company already has buyers and a far more detailed understanding of the market. They recognize the technical superiority of the first company’s widget — but they also know enough about the day to day reality of the widget world to know why it won’t sell in the numbers the first company needs.
When the first company is about to go bust, the second company purchases it at a knockdown price and begins to change the technically better widget into something with the real market fit.
Is it better to make correct decisions or fast decisions?
It’s a trade-off. And different companies will have different decision-making cultures.
Ultimately there are two factors which determine if a decision should be quick or slow
What is the minimum cost we could incur by making this decision and it being the wrong one?
Do we have the information anywhere in our organization — or the ability to get the information anywhere in our organization — to make the right decision?
Any time there is a question as to what decision to make, the reason that we are asking the question in the first place is that either both decisions are equally good (or equally bad), or that the individual making the decision doesn’t have enough information to make an informed judgment. In the case of a startup experimenting in a yet untrodden market, it is likely that there isn’t the information anywhere in the world to knowingly make the best decision. This is where concepts such as Eric Ries’ Lean Startup come from — and which lead to stories like those of the widget manufacturers.
To make a decision weigh the cost (and opportunity cost) of obtaining information against the cost (and opportunity cost) of the decision being wrong. If in doubt, go fast, because you probably don’t have all the information anyway.
2 Weigh the cost (and opportunity cost) of obtaining information against the cost (and opportunity cost) of the decision being wrong.
2.1 Anti-fragility is the ability to make small, inexpensive mistakes, and then learn from them so as to not make those mistakes again.
2.2 The more people a decision effects, the more expensive a decision failing is likely to be
2.3 The more people a decision effects, the less likely the person deciding will have good information about the effects on each person
From considering 1 and 2 above, it is clear that passing even large minority decisions up the chain of the command leads to economic disaster. The lower down a chain of command a decision is made, the less costly mistakes will be — as it will likely affect fewer people — and the more information about those it will effect will be known the decider, as they will work more closely with them.
At first glance one might think our conclusion would be to delegate all decisions. But the reality is more complicated:
James is an engineering lead. He has got to his senior position due both his unmatched skill as an engineer, but also his ability to work with people. He is rightly lauded by his superiors for building an effective team who are capable of meeting any challenge given to them. When the team has problems, James is always ready to find a solution. He works with his team members to identify their precise needs and their precise capabilities and guides the team to the correct solution. Most recently James’ team have spent time crafting a new build system — and are demonstrating to much acclaim the vast productivity improvements it has led to.
James clearly made the right decision. Didn’t he?
Well, no. What James didn’t know was that his team was about to be reorganised and merged with another team. That team have an incompatible build system — but one which many other parts of the company rely on. James ruled out moving his team to this build system (it would offer little improvement, and be a huge cost), but following the merger, his team will be paying some of those costs anyway, and experiencing new costs of working in two incompatible worlds.
As far as James was concerned, his decision was correct.
His team wholehearted agreed.
From a technical standpoint, James’ decision was excellent
Yet it was a bad decision — because it was made without all the information which was going to affect it.
It is hard to blame James for making the wrong decision. His team had never been merged before, and he had heard no rumours of such a merge — his superiors had kept the information from the team in order to not distract them from heir present work… and the superior had no information about the problems two incompatible build systems might cause.
Yet there was clearly a problem. Specifically, the problem was that James was put in a situation where he could be making decisions without enough information, and where his managers could be making the decision not to pass the necessary information on to James.
There are costs to conveying information. And these costs increase rapidly the further away from the decision maker the information is. Moreover as information is conveyed between people, it has a tendency to get confused, much like in the game of telephone.
In James’ case, his manager believed that the potential cost of giving James the information about the coming restructure was greater than the cost of James’ team doing the wrong work. Similarly James found the cost of explaining to his manager the full details of what work they were doing to be higher than the potential cost to his team of doing the wrong thing — given how unlikely that wrong thing was.
In order to minimise problems such as these we need two things: Transparency and alignment.
Transparency is the process of moving information through the company such that decisions can be made as close to where they are needed as possible, and alignment is the process of ensuring that there is a clear objective to base decision making on.
In James’ case there was insufficient transparency (James did not know that he didn’t know enough to make a decision) and insufficient alignment (James did not know that his objectives of improving the build system were not aligned with his manager’s objective of merging two teams).
It is not always possible to ensure complete alignment of objectives (since some information must legally be withheld from others), but it is valuable to ensure there is enough transparency so that it is possible to detect when objectives are misaligned.
3 Convey relevant information to as close to where decisions will be made as possible
3.1 But do not convey irrelevant information — that has a cost for not benefit
3.2 Where information cannot be conveyed to the decision making point, spend additional resource on understanding if objectives are aligned.
How does one know if there is value in conveying information?
There already exists a complicated network system which needs to have — and indeed has — solved this problem. We would do well to learn from it.
The brain is constantly receiving inputs from the various senses. However the brain is not able to manage decision making based on all of the information it received. It receives far too much to efficiently process the effects of it all. Rather it works on a somewhat upended model.
At any given time the brain is predicting the most likely things to happen next: Perhaps you hear a rustle in some leaves whilst walking through the woods. Some people, brought up in countries where this is common, would have trained their brains to predict that this rustle implies there could be a snake present. They would, based on this prediction, unconsciously prepare to act: their heart rate would increase, and they would begin to take greater notice of their surroundings. Others — like me — who have never lived in an environment where snake attacks are likely may interpret the same sounds as those of a bird, or even the wind. Our brains, based on no prediction of harm would bring about no physical changes to our body. If a snake was lurking in the woods in England, I would be unlikely to notice it, whereas someone brought up from a snake-filled environment in those same woods would be much better prepared.
Of course, since there is little risk to me from snakes most of the time, my brain can continue doing more important things… so usually this is the best action.
My brain — predicting no snake — filters out the irrelevant environment and continues with relevant thoughts about where I’m going, and how to get there most efficiently. This is a key job of the brain — not so much to determine what do do based on sensory input, but rather to determine how much sensory input should be filtered away, in order that only the most valuable data needs to be processed.
Information theory sums this up as follows: The amount of information a channel has is inversely equal to how easy it is to predict what is going to be sent through the channel. If I can predict with 100% accuracy what you are going to say to me, then I receive no information when you tell me something.
To get the most value from communication we need to be inclined to communicate most when the thing we communicate will surprise the person we are talking to, and least when they will not be surprised. In order to do this, we need to be able to understand the model the person we are communicating with uses to predict the world.
Culture (local, or corporate) comes in very handy here. Not only is culture a shared set of habits, but a shared understanding of how the world works (and thus what will surprise us about the world). When I estimate how long a project will take, a shared understanding of previous projects and estimations leaves my manager and I both knowing that I might be right, plus or minus a few weeks. If I suddenly find the project will take much more or less time, that is something I need to communicate this fact urgently. If I find I’m only going to be a few hours off in my estimation, that probably doesn’t need to be said. If there is a deadline we both know I need to meet — and new information now means that deadline will not be met, that’s a good reason to communicate. If the deadline is being moved forwards and the moving of deadlines isn’t a frequent occurrence — that’s something I need to know — even if my original estimate suggested I would still complete by the new date. Meanwhile, if a team member is ill — but another team member can fill it and not cause anything to slip, there is no need to communicate that… unless my manager asks if that team member is able to assist them with something.
It is therefore important to be clear with others what assumptions, outside of the culture you share, that you are making about the world — because it is those assumptions which will define what needs to be communicated. It is a common anti-pattern in communication that two people from different teams, different sites, or different geographies do not make it clear what they need to be told about, because in their team, site or geographic culture everybody understands such matters.
4 Communicate information which differs from what the person you are communicating expects
4.1 Do not communicate information which is irrelevant to the person you are communicating with
4.2 Ensure people you expect to communicate to you know what is relevant to you
4.3 Ensure people you expect to communicate to you know how you make predictions (and what those predictions are)
So far, the economic questions discussed have been those of authority to make a decision. However, such authority can be delegated. What is much harder to delegate is influence. In a heavily siloed company, a person may be in the ideal situation to make a policy decision, but their reach may not stretch far enough to encourage or enforce its adoption. Similarly a senior manager may attempt to enact a change in corporate behaviour — but without hands on the ground at each remote site, the desired change may remain just a decision that such a change can occur.
There is no economic value in making a decision without that decision also being acted upon. The value of a decision is thus the correctness of the decision multiplied by the effectiveness of the resulting action.
Sometimes there is no cost to obtaining the desired action. It is not unusual for a team to know they have two plausible courses of action — both of which they can support — but not to have enough information to know which path to tread. In that case (ignoring the fact that the information they need could probably have been better distributed) action will result whatever the decision.
In other extreme cases, a decision may fly in the face of a particular culture (and in this context, I am talking about culture as the embedded desires on the ground in a certain site, group or team, not the lofty desires of a remote management summed up in a brief and pithy values document). In this case, people will actively resist the decision, as it will go against all the behaviours their group has built itself upon. In the latter case one must consider
“Does the value of making this decision outweigh the costs of imposing it?”
How one might impose such a decision is outside the scope of this essay — but it may well involve monitoring, costly actions, over-communication and building a team of influencers. It is clear that this could be mightily costly — so the benefits would have to be huge. This would also be a case where one would need to be certain the decision was correct, as reversing it would still involve incurring the costs accumulate up to the point of reversal — and might leave additional cultural costs for the organisation to pay back over months and years to come.
One must, however, notice that the location within the organisation where the decision is made and the location where the influence comes from do not need to be the same. The influencers in an organisation need not be the decision makers, and vice versa. Indeed the king and vizier model has historically been shown to be an effective pattern for running everything from teams to states.
Influence has a different set of economics from decision making, but there are a few rules which can guide the decision maker in minimising the costs of influence — and thus maximising their options when it comes to decision making
Establish a culture of supporting decision makers: When a decision has to be made, it is frequently better to have a plausible decision being tested than the one absolutely correct decision being discovered and argued about through a lengthy and costly process. For a team which will need to jointly be responsible for implementing whichever decision is made, you should adopt a policy of ‘decide and commit’ — whoever is given the ultimate responsibility of making the decision, the entire team should agree on the decision maker’s authority and agree that — whatever their opinion — they will commit to it.
Commit In public : decisions will often be passed down from ‘on high’ but be implemented on the ground. This is frequently necessary in order to exploit economies of scale within an organisation. Many people in the middle of the hierarchical chain may find they disagree with the policy — and believe their circumstances — or the circumstances of their team — have not been considered. The culture of optimum decision implementation suggests such people should be able to dissent (or inform, or question) upwards, in order to make sure that the information they have reaches the decision maker — and also to ensure they are able to learn the information which informed the decision maker, however downwards and outwards, while they may display empathy for the problems that the decision may cause, they should also enforce and advocate for it.
Allow an ‘intend culture’ rather than a ‘request culture’ to build ownership. If you are relying on members of your team — especially those without wide influence — to make decisions, you need to empower them to be able to make the decisions. This may well involve sharing your own influence as a backer of those decisions. This can be tricky, as it might involve backing a decision where you are not fully informed… or worse backing a decision where you might have information which, if shared, might have changed the decisions outcome.
In his book ‘Turn the Ship Around’ David Marquet talks about a culture he both experienced and later implemented which allowed him to be better able to make these kinds of decision. As a submarine commander, he was used to a strict command and control environment where crew asked permission to perform actions, and he would decide upon whether or not they should go ahead. Marquet shifted this slightly by instead getting his crew to tell him what they intended to do. A subtle word change, but a change which shifted the decision making process. Marquet could then question the decision (a clever strategy he used was to ask “what questions do you think I should be asking you?”) and could take the opportunity to provide more information, if he felt it might be needed, but he could rely on his crew’s expertise to ultimately do the right thing.
Once informed of an intended action and happy with the reasoning behind it, you can commit to supporting it without needing to be the decision maker — or even have all the information needed to make the decision.
5 Empower your team to make decisions through an intend rather than request culture
5.1 Empower your team to empower their team to make decisions
5.2 Commit to decisions — whoever the decision maker has been empowered to be
5.2.1 Support your team’s decisions — and provide influence to back your team’s decisions
5.2.2 When decisions have to be made and there is no consensus, agree on the decision maker and agree to commit to the decisions her make
5.3 Ensure decisions are communicated to all the relevant stakeholders — that is everyone affected by the decision
5.3.1 Ensure the communication comes with the required influence
In considering the economics of decision making and their corollaries as listed above it is clear that ultimately there are two factors which economically efficient decision making hangs on:
Economic communication
and
Optimising for speed of decision making over accuracy of decisions
Such principles find themselves reflected in lean and agile work processes — all the way back to the Toyota Manufacturing System, in systems of alignment such as Andy Grove’s OKRs and in the day to day organisational structures of efficient self-organising groups such as many open source software projects. Nevertheless command and control is inherently wired into our minds, and it can be hard to resist — especially when traditional organisational structures reinforce it. To act efficiently, you need to group people by the information they need, not by their role or skills. Furthermore, such groups need to be co-located to ensure the lowest cost of communications. Optimising for decision making is not something you can impose on top of an existing organisation structure, it is something that an organisation must build — or rebuild — itself around.
This too is a cost, but the potential results are a dynamic system which supports rather than frustrates the organisation’s goal of producing value for all stakeholders.